Angle of attack measurement using low-cost 3D printed five hole probe for UAV applications

Abstract This paper presents a low-cost 3D printed Five Hole Probe (FHP) for Angle of Attack (AOA) measurement and a solution to predict the missing pressure data due to hole blockage using machine learning. The five Hole Probe is designed and fabricated taking reference to the open-source probe called “the oxford probe”. It is intended to use this probe in Mini Unmanned Aerial Vehicles for AOA measurement because the UAVs are primarily designed to carry out unconventional missions like flying at low speed and low altitude, which makes the AOA measurement as one of the absolute requirement. The main challenges involved in mini flyers are their size and weight constraints. Due to these constraints, the probe should have lightweight but highly accurate. Keeping this in mind, suitable design modifications and material selection are made on the Oxford probe and it is manufactured by 3D printing. Then the 3D printed probe is tested in the wind tunnel. In parallel, a CFD analysis of the FHP is carried out in the ANSYS WORKBENCH environment and the results are presented. The CFD results are compared with windtunnel measurements of AOA, and the results are analyzed. Calibration of FHP is carried out, and a lookup table is generated using the pressure coefficients C P α and C P β corresponding to respective AOA and AOS. For validation, the probe is kept at known AOA and AOS, and the pressure coefficients were calculated. It has been found that the accuracy of measurement is increased by using the mean of ten pressure samples for calculating the pressure coefficients, instead of every pressure sample. And due to the small diameter, the ports may get blocked from dust particles. The machine learning algorithm (SVR – Support Vector Regression) has been trained and tested to tackle the problem of hole blockage. Various regression models were tested to predict the missing Pressure. The Quadratic SVR regression model is selected based on RMSE value. The selected regression model is validated by blocking one of the ports of FHP.